[PATCH] Debugging LLVM IR with LLDB (and gdb)

[Andrew Trick]

On May 8, 2013, at 2:04 PM, "Malea, Daniel" <daniel.malea at intel.com> wrote:

> Thanks for the reviews guys.
> 
> I committed in 4 stages; r181463, r181466, r181467 and r181469.
> 
> The tests had to be re-generated since my initial implementation as they
> contained debug metadata which must be parsed by DebugInfoFinder (due to a
> pesky limitation of my current implementation) and it seems the debug
> information format for !llvm.dbg.cu has been slightly changed in the last
> month.
> 
> I expect they'll break as well the next time that debug information is
> updated. As a temporary workaround, shall I also check in the original
> (C/C++) sources, as well as the script that generates the tests?
> 
> I hope to find some time soon to "finish" the implementation and not
> require existing debug info from "clang -g" or equivalent, but for now, I
> think checking in the source code would be best. At the same time, I know
> there's been some concerns in the past with checking in C/C++ code into
> the LLVM 'test' directory.. So I figured I'd ask before doing so.

I imagine that c/c++ code needs to be in clang/test.

More importantly though, the .ll input should be free of metadata. I suggest focussing on fixing that before the debug format changes again.

There is still the issue of your CHECKed output matching the current format. But in that case, when the test breaks, someone can just copy the new output in to the CHECK file. Right?

-Andy

> On 2013-05-07 4:31 PM, "Eric Christopher" <echristo at gmail.com> wrote:
> 
>>> Meanwhile, I think you should commit what you have.
>>> 
>> 
>> Agreed.
>> 
>> -eric

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130508/e3e531c7/attachment.html>